Distributed broadcasting system

This repository presents a fast, fault-tolerant distributed broadcast system that allows you to broadcast messages to multiple nodes on any network topology. Configurable for different loads and different topologies. A message that was sent to one node will quickly appear on all the others and temporary network problems will not lead to never-read or duplicated messages, due to asynchronous synchronization.
The protocol is based on the maelstrom protocol and runs only with this protocol.

Performance

For grid-topology on 25 nodes, with 100 messages per second and 100ms network latency, a result of 12 messages per operation is achieved with median latency < 1 second and worst latency < 2 seconds (~1600 ms) (syncDeadline should be 30 ms to achieve this)

There is a chance to achieve a better performance for different loads by configuration of node and syncer, while keeping same code.

For testing with maelstrom:

./maelstrom test -w broadcast --bin <your-binary-path> --node-count 25 --time-limit 20 --rate 100 --latency 100

Protocol brief description

Each node is equal. Each node receives messages from the network either from outside or from other nodes

• broadcast a message containing data that should be on other nodes
• read a message to which the node must respond with all the messages it has received on the network
• topology is a message that defines the neighbors of a node - the addresses of the node with which it can communicate over the network
• sync is a message used for asynchronous synchronization between neighboring nodes